• Study the molecular basis of variation in development and disease
• Using high-throughput experimental methods

• For over a decade we have laid the basic molecular blueprint by sequencing DNA

build a whole human genome sequencing device and use it to sequence 100 human genomes within 30 days or less, with an accuracy of no more than one error in every 1,000,000 bases sequenced, with an accuracy rate of at least 98% of the genome, and at a recurring cost of no more than $1,000 (US) per genome.

“genome sequencing technology is plummeting in cost and increasing in speed independent of our competition”

“companies can do this for less than $5,000 per genome, in a few days or less — and are moving quickly towards the goals we set for the prize.”

NHGRI strategic plan

NHGRI strategic plan

"The major bottleneck in genome sequencing is no longer data generation—the computational challenges around data analysis, display and integration are now rate limiting. New approaches and methods are required to meet these challenges."

• Data analysis
• Data integration
• Visualization
• Computational tools and infrastructure

What makes them different?

Much human variation is due to difference in ~ 6 million DNA base pairs (0.1 % of genome)

What makes them different?

Same genome is expressed differently during different stages and in different tissues.

DNA is packed, making parts inaccessible, and this packing is dynamic!

DNA methylation is a chemical modification of DNA, regulates gene expression.

Measuring DNA methylation and understanding role in expression regulation in solid tumors

Large blocks of hypo-methylation (sometimes Mbps long) in colon cancer

• Hansen, et al., Nat. Genetics, 2011
• Corrada Bravo, et al., BMC Bioinformatics, 2012
• Timp, et al., Genome Medicine, 2014
• Dinalankara and Corrada Bravo, Cancer Informatics, in press.

Genes with hyper-variable expression in colon cancer are enriched within these blocks.

• Hansen, et al., Nat. Genetics, 2011
• Corrada Bravo, et al., BMC Bioinformatics, 2012
• Timp, et al., Genome Medicine, 2014
• Dinalankara and Corrada Bravo, Cancer Informatics, in press.

Genes with consistent hyper-variable expression across tumors are tissue-specific.

• Hansen, et al., Nat. Genetics, 2011
• Corrada Bravo, et al., BMC Bioinformatics, 2012
• Timp, et al., Genome Medicine, 2014
• Dinalankara and Corrada Bravo, Cancer Informatics, in press.

Motivated by observations that gene expression hyper-variability is enriched in specific regions of epigenetic alteration in colon cancer…

…and that consistent hypervariability across tumor types is enriched in genes involved in tissue specificity

anti-profile score: measures sample-specific deviation from normal expression in consistently hyper-variable genes

Good cross-experiment properties
Stability in normal expression across experiments

Prediction in leave-one-tissue out experiment

Anti-profile score distinguishes between stages in tumor progression

DNA methylation anti-profiles score distinguishes between stages in tumor progression

Stratification based on anti-profile score

• Simple counting scheme produces robust stable and accurate (anti)-profiles
  
• Nice prediction properties across experiments and across tissue types
  
• Captures increasing hyper-variability associated with progression and prognosis

Support Vector Machines for Anomaly Detection: determine if observations belong to a given group or are anomalies.

• Distinguish observations from two anomalous groups (e.g., adenoma vs. tumor)

• How can we incorporate the fact that we are classifying anomalies?

• Why (and when) is it worth doing that?

• Based on leave-one-out error bounds (via stability arguments)
• \(K \succeq \tilde{K}\), implies LOO error bound of Anomaly SVM is lower than LOO error bound of standard pairwise between anomalies SVM
• Proof uses arguments based on the SVM path algorithm, and required conditions

Prediction of high vs. low relapse risk in lung cancer

Prediction of suspect vs. pathological fetal CTG data (not genomics)

• Profiles learned based on hyper-variability show consistent behavior across tissues and across experiments in tumor prognosis and progression

• We can extend the general anti-profile idea to a function approximation setting

• Use sensitivity-based cross-validation error bounds to characterize the effect of incorporating normal observations when classifying between anomalies

• Indirect similarity through normal samples improves stability while improving prediction performance

• State-of-the-art computational and statistical analysis platform
• We develop and apply methods for these analyses in this platform
• Our collaborators take part in analysis in this platform

• Data transformation and modeling: data smoothing, region finding (R/Bioconductor: Bsmooth, minfi)
• Genome browsing: search by gene, search by overlap
• Region analysis: overlap with other data (our own, other labs, UCSC, ensembl)
• Regulation: expression data (Gene Expression Barcode)

• We have entered the era of data analysis
• Epigenomics roadmap, Encode, TCGA, …
• We have unprecedented ability to measure
• and lots of publicly available data to contextualize it

Integrative, visual and computational exploratory analysis of genomic data

Florin Chelaru, UMD

Chelaru, et al. Nature Methods, 2014.

The old way

Integrative, visual and computational exploratory analysis of genomic data

• Browser-based
• Interactive
• Integration of data
• Reproducible dissemination
• Communication with R/Bioconductor: epivizr package

Using the epivizr package

• Setup up an epivizr session

mgr <- startEpiviz(workspace="qyOTB6vVnff")

• Calculate a statistic of interest

# Get tumor methylation base-pair data
m <- assay(se)[,"tumor"]

# Compute regions with highest variability across cpgs
region_stat <- calcWindowStat(m, step=25, window=80, stat=rowSds)
s <- region_stat[,"stat"]

Using the epivizr package: browse by regions of interest.

• What's around the regions with highest across CpG variability

# get locations in decreasing order
o <- order(s, decreasing=TRUE)
indices <- region_stat[o, "indices"]
slideShowRegions <- rowRanges(se)[indices] + 1250000L
mgr$slideshow(slideShowRegions)

Our architecture is dynamically extensible. We can easily integrate new data types and add new visualizations.

Example: adding a new visualization

see: https://gist.github.com/11017650

http://epiviz.cbcb.umd.edu/?gist[]=11017650&ws=Y8kWxCO2Ajn

http://epiviz.cbcb.umd.edu/?ws=SRHZlWRRAPd&gist[]=a82a998817564ce3fe48&settings=default&

One interpretation of Big Data is Many relevant sources of contextual data

• Easily access/integrate contextual data
• Driven by exploratory analysis of immediate data
• Iterative process
• Visual and computational exploration go hand in hand

• Context
• Integrate and align multiple data sources; navigate; search
• Connect: brushing
• Encode: map visualization properties to data on the fly
• Reconfigure: multiple views of the same data

• Data
• Select and filter: tight-knit integration with R/Bioconductor;
• (future) filters on visualization propagate to data environment
• Model
• New 'measurements' the result of modeling; perhaps suggested by data context

We are building a software system to support creative exploratory analysis of epigenome-wide datasets…

Computed Measurements: create new measurements from integrated measurements and visualize

Summarization on the fly: create new measurements from integrated measurements and visualize

Beyond genomics and epigenomics: metagenomics

Beyond genomics and epigenomics: metagenomics

Coordinates:

Beyond genomics and epigenomics: metagenomics

Beyond genomics and epigenomics: metagenomics

Beyond genomics and epigenomics: metagenomics

Acknowledgements

Wikum Dinalankara: (formerly) Ph.D. student @ U. Maryland (now postdoc @ Johns Hopkins)
Nick Thieme: Ph.D. student @ U. Maryland
Jeff Leek, Rafael A. Irizarry, Andy Feinberg

References

Corrada Bravo, et al. (2012) BMC Bioinformatics.
Dinlanakara et al. (2015) Cancer Informatics.
Dinalankara et al. In preparation

Contact

http://hcbravo.org
@hcorrada